[time] garmin gps25-hvs questions
gnu not unix
Sat Dec 4 17:24:07 UTC 2004
In message <nvl3r0tqr0qcu41lnedimvo13qudm6easn at 4ax.com> you write:
>worrying about calibrating for cable delays, etc. It is a Garmin
>GPS25-HVS model (6-40 volt, rs232 output).
I have a gps35-lvs and it has worked admirably here, as a
reference clock working with ntpd.
>1) how do I set the baud rate to 4800 on /dev/gps1? or /dev/ttyS1
>(machine is Debian Linux) Or do I not need to, is it done
>automatically by the ntp refclock driver?
The ntpd software has a fixed 4800 baud configuation. This matches
a typical NMEA gps unit such as the gps25. It works fine and
is set by ntpd when the software starts up.
>2) Do I have to do anything special to the PPS output line in terms of
>buffering, or do I just connect to the DCD pin? or some other pin?
Yes, the PPS TTL level signal is not compatible with rs232.
A MAX232 TTL to rs232 converter runs on +5 volts and will
make the PPS compatible with the DCD pin on a pc's serial port.
You will need a 7805 style +5 volt regulator to power this.
>3) If I connect all the unused pins on the serial port together via
>diodes, is there enough power available to run the GPS rx (1W, 80mA)?
>Or do I need to temporarily run it off a 12v sealed lead/acid battery
>I have lying around?
I'd not try sucking more than a few milliamps out of rs232 pins.
Wire up a small fuse and connect to your pc's +12 rail. A ferrite
core might help reduce possible spurious radiation.
There are more details about PC's and ntp at this web site:
Basically, if your motherboard is decent, you should be able
to achieve single digits microsecond performance. This would be a fine
stratum 1 addition to the pool.ntp.org herd.
Note that there are a fair number of things to verify, before
your setup can achieve this level of performance. ACPI, spread
spectrum, serial latency, oscillator drift, and unix kernel
configuration are all items that must be taken into consideration.
More information about the pool